Low birth weight is a major public health problem and birth weight has been the subject of numerous epidemiologic studies. Infant birth weight has almost exclusively been abstracted from medical records or birth certificates, and there has been little attempt to assess the accuracy of maternal recall of birth weight. In this project, the birth weight of a mother's first child, as stated by the mother at registration for prenatal care during her next pregnancy, was compared to the birth weight as given in that child's medical records. There were 7521 pregnancies studied. Overall, only 4.2% of women were unable to state a birth weight for their most recently born child. Blacks, women of higher parity, younger women, and women with less education were less likely to state a birth weight for their most recently born child. Mothers were less likely to state a birth weight for infants less than 1000 grams and infants who did not survive the first 48 hours. Among those who did state a birth weight (n=5582) the errors in birth weight was unbiased with respect to most factors, but there was a slight tendency to overstate the birth weight of small infants and understate large ones. Low birth weight was identified quite accurately. The sensitivity of reported low birth weight was 96%, and the specificity of reported normal birth weight was 98%. It is concluded that among reproductive age women, recalled birth weight can reasonably be substituted for medically verified birth weight.